Sui Foundation Explains Why the Mainnet Stopped 3 Times in 2 Days

A single upgrade introduced two separate bugs that cascaded into three outages within 48 hours. Here is what actually went wrong.

One Upgrade, Two Bugs, Three Outages

The Sui mainnet went down three times in 48 hours last week, and the Sui Foundation has now published a detailed explanation of how it happened. The short version is that a single upgrade, version 1.72, introduced a new feature that interacted badly with existing gas-charging logic in a way nobody caught before deployment. What followed was a sequence of failures where each fix created the conditions for the next problem.

The upgrade itself introduced something called Address Balances, a new way for users to store funds and pay transaction fees without using traditional coin objects. On paper it was an improvement. In practice it exposed an edge case in how the network handles canceled transactions.

The Gas Bug That Started Everything

When two transactions compete to spend from the same address balance simultaneously and there are not enough funds to cover both, the network cancels one of them. That is the intended behavior. The bug was in what happened next. Even after a transaction was marked as canceled due to insufficient funds, the gas-charging process still attempted to debit those same funds during a step called gas smashing, which combines multiple payment sources before charging fees.

The result was a negative balance where zero was expected, and that crashed the validator nodes. The first outage began Thursday morning and lasted roughly six and a half hours before enough validators adopted an emergency patch to bring the network back.

The emergency patch was not a complete solution. The team acknowledged at the time that it carried a small but real probability of triggering the same crash under slightly different conditions, specifically if a transaction had multiple cancellation reasons and one masked the other. They accepted that risk in order to restore the network quickly while working on a permanent fix. Friday morning, that exact scenario materialized and the network went down again. The second outage lasted around three and a half hours.

The Third Problem Was Different

By Friday morning the core team had a robust fix ready. Validators restarted to install it, and the network came back up. What nobody anticipated was that the restart cycle would create the conditions for a third, entirely separate failure later that afternoon.

Sui’s network uses a distributed key generation protocol at the start of each epoch to establish the randomness that certain transactions depend on. That protocol requires a minimum level of validator participation to run. When validators restarted to adopt the Friday morning fix, participation temporarily dropped below that threshold, and the randomness system automatically disabled itself. That part worked as designed.

The bug was in what happened after. The disabled status was never written to disk. When validators came back online they had no record that randomness had been turned off, so they continued expecting it to function. Transactions that required randomness began queuing up waiting for a system that was effectively not there. The end-of-epoch process, which must clear that queue before it can close, sat waiting indefinitely. The third outage lasted nearly six hours.

How It Was Fixed and What Changes

The permanent fix addressed both the gas-charging underflow and the randomness state persistence problem. An additional mechanism was added that allows validators to force-close a stuck epoch at a coordinated point, which was used once to exit the affected epoch and restore normal operation. The Foundation confirmed no user funds were at risk at any point and no previously confirmed transactions were rolled back.

The post-mortem also noted that AI agents with access to validator logs and production metrics meaningfully accelerated the diagnosis process during the incident, broadening which engineers could actively debug the live network.

Now the team identified three areas requiring deeper investment: end-of-epoch resilience, gas-charging code quality, and better failure containment so that a crashing input can be isolated rather than bringing down the entire network. The framing is honest about the fact that gas-charging logic has grown complex enough that edge cases like these are difficult to rule out through code review alone, and that the systems designed to protect the network need to be hardened before the next upgrade cycle.

How the price reacted

Despite the network being offline for roughly 16 hours across the two days, the price damage extended well beyond the outage windows. SUI dropped 7.27% on May 28th alone, the day the first halt began, and the selling continued through the weekend. Price is now trading at $0.876, having broken below two significant levels simultaneously.

The 0.786 Fibonacci at $0.927 and the 100 SMA at $0.965 both gave way, stripping away the support structure that had held since the April recovery. Adding to that, price has also broken below an ascending trendline, a rising support line that had been intact since the March lows. Losing all three at once leaves SUI in technically vulnerable territory.

The next reference on the chart is the 1.0 Fibonacci at $0.795, a level SUI has not traded at since early February. The 50 SMA at $1.003 and 100 SMA at $0.965 now sit above as resistance. RSI at 34.51 is pressing toward oversold territory also last seen during the February lows, which could slow the selling momentum from here.

The information provided in this article is for educational purposes only and does not constitute financial, investment, or trading advice. Coindoo.com does not endorse or recommend any specific investment strategy or cryptocurrency. Always conduct your own research and consult with a licensed financial advisor before making any investment decisions.

Next article